Hepatic dysfunction and thrombocytopenia induced by excess sFlt1 in mice lacking endothelial nitric oxide synthase

Liver dysfunction is a major problem in patients with severe preeclampsia (PE), hemolysis, elevated liver enzymes, and low platelet count (HELLP) syndrome, or in patients receiving anti-vascular endothelial growth factor (VEGF) therapy. Excessive soluble fms-like tyrosine kinase 1 (sFlt1) that antagonizes VEGF has been implicated in the pathogenesis of PE. VEGF increases the expression of endothelial nitric oxide synthase (eNOS) and activates it. eNOS polymorphisms that cause reduced NO production are associated with PE. The aim of this study was to clarify the role on hepatic function by excess sFlt1 in the absence of eNOS gene product. We first overexpressed sFlt1 using adenovirus in eNOS −/− and eNOS +/+ mice. Excessive sFlt1 and lack of eNOS synergistically increased plasma levels of liver transaminases, exacerbated infiltration of inflammatory cells, elevated expression levels of cytokines in the liver, and aggravated oxidative stress and coagulation abnormalities. Lack of eNOS in the presence of excess sFlt1 also induced thrombocytopenia, whereas eNOS +/+ mice with excess sFlt1 alone showed no or modest liver phenotype. Taken together, excessive sFlt1 and lack of eNOS synergistically induce hepatic dysfunction and thrombocytopenia, suggesting a novel role for VEGF and nitric oxide signaling in hepatocyte-endothelial cross-talk in health and in liver injury states.

demonstrated that lack of eNOS exacerbates sFlt1-induced kidney injury through endothelin activation 15 . Based on these findings, we hypothesized that eNOS dysfunction is likely involved in the exacerbation of tissue injury caused by VEGF inhibition.
Here, we demonstrate that excessive sFlt1 combined with lack of eNOS in non-pregnant mice causes severe liver dysfunction accompanied by hepatic inflammation, oxidative stress, and dyslipidemia. Coagulation abnormalities and thrombocytopenia were also evident.
Histological damage and inflammation in the liver. Since lack of eNOS in mice with excessive sFlt1 further increased the levels of ALT and AST, we next performed pathological analysis of the liver. Figure 2a-c shows the representative hepatic photomicrographs of Hematoxylin-Eosin stain, TdT-mediated dUTP nick end labeling (TUNEL) stain and immunohistochemistry against cleaved caspase 3. The liver from eNOS −/− ; sFlt1 mice exhibited hepatocyte ballooning accompanied by vacuolar degeneration, necrotic lesion, and infiltration of inflammatory cells (Fig. 2a). The number of inflammatory foci was significantly increased in eNOS −/− ; sFlt1 mice compared to that of control and eNOS +/+ ; sFlt1 mice ( Fig. 2d). There was an increase in the hepatocyte ballooning score with excessive sFlt1 in both eNOS +/+ mice and eNOS −/− mice (Fig. 2e). The cleaved caspase 3 positive hepatocytes were frequently observed in the liver from eNOS −/− ; sFlt1 mice ( Fig. 2f). To further analyze the inflammation in the liver, we tested the changes in neutrophil infiltration and the expression of proinflammatory and profibrotic genes in the liver. Excessive sFlt1 in the eNOS −/− mice increased the number of infiltrating neutrophils (Fig. 3a,b). The levels of hepatic myeloperoxidase (Mpo) mRNA in the eNOS −/− ; sFlt1 mice were more than 500 fold higher than that of the eNOS +/+ mice (Fig. 3c), whereas excessive sFlt1 per se did not affect macrophage infiltration (Supplementary Figure 2a,b). As shown in Fig. 3d, lack of eNOS elevated the expression levels of Tnfa, Ccl2, Cxcl2 and Vcam1 only in the setting of excessive sFlt1. Similarly, the levels of pro-fibrotic genes, Col1 and Acta2, were elevated in the eNOS −/− ; sFlt1 mice (Fig. 3d). These findings indicate that in the setting of excessive sFlt1, lack of eNOS exacerbates histological damage and inflammation in the liver.
Oxidative stress and hypoxia. Inhibition of VEGF or of eNOS exacerbates hepatic hypoxia pathways 2,16 , which increases oxidative stress and promotes liver injury 17 . Accordingly, we examined oxidative stress and hypoxia in this model. Intensity of immunoreactive 4-hydroxy-2-nonenal (4HNE) in the liver from the eNOS −/− ; sFlt1 mice was significantly higher than that from the eNOS +/+ mice with or without excessive sFlt1, suggesting that oxidative stress is increased in the liver from the eNOS −/− ; sFlt1 mice ( Fig. 4a,b). The protein and gene expression of HO-1 (Hmox1), an anti-oxidative enzyme, was increased in the liver from the eNOS −/− ; sFlt1 mice ( Fig. 4c-e). Similarly, gene expression of Nqo1 was up-regulated (Fig. 4f). Strong immunoreactive hypoxia inducible factor 1α (HIF1α) was observed in the liver from the eNOS −/− ; sFlt1 mice ( Fig. 4g). Moreover, the gene expression of Glut1 and Epo, other typical target genes of HIF, was significantly upregulated in the eNOS −/− ; sFlt1 mice (Supplementary Figure 3a,b). We conclude that lack of eNOS in mice with excessive sFlt1 aggravates oxidative stress and hypoxia, which likely induces severe liver injury.  Thrombocytopenia induced by excessive sFlt1 in mice lacking eNOS. Because VEGF inhibitor therapies and severe preeclampsia is characterized by hematological abnormalities 20,21 , we next examined hematological parameters in this model. The number of fibrin thrombi was significantly higher in the liver from the eNOS −/− ; sFlt1 mice than that from other three groups of mice (Fig. 6a,b). Complete blood count showed reduced platelet number and increased white blood cell number in the eNOS −/− ; sFlt1 mice ( Fig. 6c,d). The red blood cell count and hematocrit were similar among the groups (Fig. 6e-g). These findings indicate that lack of eNOS in mice with excessive sFlt1 caused hypercoagulability and thrombocytopenia. Although inhibiting VEGF causes thrombotic microangiopathy and hemolytic anemia 22,23 , excessive sFlt1 and lack of eNOS did not cause anemia. Consistent with this observation, the level of plasma haptoglobin, a marker of hemolysis, was not statistically different between the groups (Supplementary Figure 4). Moreover, schistocytes were not observed in their smears (data not shown). We conclude that lack of eNOS in the context of excessive sFlt1 exacerbates hypercoagulability and thrombocytopenia without obvious hemolysis.

Discussion
We have demonstrated that the lack of eNOS in the presence of excessive sFlt1 exacerbates hepatic injury and causes hypercoagulability and thrombocytopenia. Our data show that the livers from the eNOS −/− ; sFlt1 mice have enhanced hepatic inflammation, prominent neutrophil infiltration, and increased oxidative stress and the expression of genes induced by hypoxia. Literature shows that nitric oxide (NO) derived from eNOS is anti-inflammatory in vitro and in vivo 16,[24][25][26] . NO-donor directly reduces the expression levels of hypoxia-induced cytokines and chemokines in HepG2 cells 24 . Lack or inhibition of eNOS exacerbates hepatic inflammation in obesity and ischemic models 16,25,26 . Moreover, hepatocyte specific deletion of VEGF causes hypoxia 2,27 , which causes tissue injury mediated by HIF 28 . Up-regulated HIF1α and HIF2α increase hepatic inflammation, and contribute to alcoholic or non-alcoholic liver disease and acetaminophen induced liver injury [29][30][31] . Consistent with these findings, our data suggest that inflammation, hypoxia, and oxidative stress could be an important pathogenic factor in the exacerbation of liver injury in setting of reduced VEGF signaling and impaired NO production.
In our experimental condition, adenovirus increased plasma sFlt1 concentration to ~1.0 × 10 4 ng/ml (Supplementary Figure 5). Previous report demonstrated that such a high level of sFlt1 almost completely inhibits VEGF signaling 10 . Furthermore, the lack of any phenotype in the control adenoviral group suggests that the phenotype induced by sFlt1 is specific to VEGF inhibition. Moreover, eNOS dysfunction is likely crucial to the onset or exacerbation of VEGF inhibitor-induced liver injury, because wild type eNOS mice with extremely excessive sFlt1 did not show hepatic damage. The patients with PE and HELLP syndrome have elevated levels of serum triglyceride and fatty acid compared to those of normal pregnancy 18,19,32 , suggesting that inhibiting VEGF is associated with abnormal lipid metabolism in the liver. In accordance with this finding, our data indicate that sFlt1 overexpression increases the levels of plasma triglyceride and total cholesterol. However, lack of eNOS did not further exacerbate these parameters (Fig. 5). Literature shows that hepatocyte specific inhibition of triglyceride-rich lipoprotein clearance receptors Ldlr or Lrp, elevated plasma lipoprotein 33 , and that skeletal muscle and adipose tissue actively regulates lipoprotein clearance 34 . Consistent with these findings, excessive sFlt1 reduced the expression levels of Ldlr and Lrp1 in the liver (Fig. 5d). Despite liver damage and abnormal lipid profile in the plasma, the liver did not show increased triglyceride content in the eNOS −/− ; sFlt1 mice. It is likely that reduced fatty acid oxidation in the liver suppresses lipogenesis and lipid uptake in the liver, leading to increased plasma triglyceride levels. There was no remarkable effect of lack of eNOS on lipid metabolism in the liver.
Thrombocytopenia is a characteristic feature of VEGF inhibitor-induced thrombotic microangiopathy 22 , but its pathogenesis remains unclear. Previous report demonstrates that excessive sFlt1 together with lack of Adamts13 develops hemolysis and thrombocytopenia in mice 35 . Pregnant mice with excessive sFlt1 and soluble endoglin mimic features of human HELLP syndrome 8 . Our data and these findings suggest that excessive sFlt1 alone is not sufficient to cause thrombocytopenia. Because NO derived from eNOS inhibits platelet activation 24,36 , we suggest that reduced NO from eNOS causes thrombocytopenia when VEGF is inhibited.
Stringent VEGF inhibition increases hepatocyte erythropoiesis and polycythemia, which is mediated by increased erythropoietin production due to HIF2 activation 10,37 . But this was not evident in our model, although the levels of Epo mRNA in the liver were elevated with excessive sFlt1 (Supplementary Figure 4b). Overexpression of both sFlt1 and soluble endoglin displays the phenotype of HELLP syndrome including hemolysis 8 . Excessive sFlt1 together with lack of eNOS is not sufficient to cause hemolysis, and overexpression of both sFlt1 and soluble endoglin is likely necessary for hemolysis to cause HELLP syndrome. However, some preeclamptic patients have liver injury and thrombocytopenia without hemolysis 38 , and our model could explain the pathogenesis of these patients.
sFlt1 is known to inhibit PlGF signaling. However, PlGF is largely made during pregnancy and at least 7-8 folds lower in non-pregnant states 39 . Moreover, lack of Plgf does not affect normal angiogenesis, and PlGF blockade rather ameliorates liver fibrosis and inflammation in cirrhotic mice [40][41][42] . Hepatic expression of PlGF is undetectably low in our preliminary observation and in prior reports 40,42 . Accordingly, we believe inhibition of PlGF does not contribute to exacerbation of liver toxicity by excessive sFlt1 in our model. However, whether this is true during pregnancy where PlGF is abundantly made needs additional studies.
We used non-pregnant mice with excessive sFlt1 because increased sFlt-1 recapitulates the phenotype of maternal syndrome of preeclampsia regardless of pregnancy in rodent models 6,15 . However, sFlt1 explains only some aspects of the pathogenesis of preeclampsia. Various factors including endoglin, endothelin, catechol-O-methyltransferase, or angiotensin-II are likely involved in endothelial dysfunction in pregnant or preeclampsia condition 8,15,43,44 . Their interaction with eNOS and the role in hepatic injury should be clarified in the future.
In conclusion, we have demonstrated that hepatotoxicity of sFlt1 is exacerbated by lack of eNOS. Further studies should evaluate the nitric oxide independent pathways induced by VEGF inhibition. These findings might open a novel role of hepatocyte-endothelial communication in the liver homeostasis and underling mechanism of liver injury induced by impaired VEGF signaling.

Methods
Animals. All experiments were conducted in compliance with the guidelines of Tohoku University.
Experimental protocol was approved by the Institutional Animal Care and Use Committee at Tohoku University. Ten to fourteen-week-old non-pregnant female eNOS −/− mice with C57BL/6 J genetic background were injected with 1 × 10 9 PFU adenovirus to overexpress sFlt1 (Adeno sFlt1) or adenovirus encoding GFP protein (Adeno GFP) at equivalent doses as we previously described 15,45 . These mice were maintained for 7 days. Previous studies have shown that increased sFlt-1 recapitulates the phenotype of preeclampsia regardless of whether the animal is pregnant 6,15,45 . Mice cannot maintain pregnancy if excessive sFlt-1 and lack of eNOS are combined (our unpublished observation) 15 . Accordingly, we used non-pregnant female eNOS −/− mice for these studies.
Biochemical measurement. ELISA kits were used to measure urinary albumin (Exocell Inc., Philadelphia, PA), plasma sFlt1 (R&D Systems Inc, Minneapolis, MN) and plasma haptoglobin (Life Diagnostics, Inc. West Chester, PA). Colorimetric detection kits were used to measure AST, ALT, triglyceride and total cholesterol (Wako chemicals, Osaka, Japan) in plasma and liver homogenate. Urinary creatinine was determined by the method we developed using LC-MS/MS 46 .
Blood count. Blood was collected with EDTA and analyzed using Microsemi LC-662 (Horiba, Japan).
Quantitative RT-PCR. Total RNA from the liver was extracted using TRI Reagent (Molecular Research Center, Inc., Cincinnati, OH). Hypoxanthine-guanine phosphoribosyltransferase (Hprt) was used as a reference gene as we previously reported 47,48 . The primers used in this study have been previously described elsewhere. Their sequences are available on request.
Morphological study. Livers were fixed in 2% PFA and embedded in paraffin. The sections 2 μm in thickness were stained with Hematoxylin-Eosin stain to evaluate histological damage. The degree of lobular inflammation was evaluated by counting of inflammatory foci. The degree of Injured hepatocytes was examined using SCIENTIFIC RepORtS | (2018) 8:102 | DOI:10.1038/s41598-017-18260-7 ballooning score as previously described 49 . Ballooning score was determined according to number of ballooned hepatocytes: 0 (none), 1 (few), and 2 (many). 5 consecutive fields were examined in each slide at 100-fold magnification. All examination was performed under blinded manner.
Statistical Analyses. Multiple groups were compared using two-way ANOVA with the Tukey-Kramer test for parametric values, if necessary logarithm transition was performed. Otherwise, Kruskal-Wallis test with Dunn's test was used for non-parametric values. All analyses were performed using JMP 11.0.0 (SAS Institute Inc., Cary, NC). Values are presented as mean ± s.e.m or box plot. Differences were considered statistically significant with P < 0.05.